Conversation
README.md
Outdated
| * [Metrics](metrics.md) | ||
| * [Installation](installation.md) | ||
| * [Usage](usage.md) | ||
| * [Configuration](documentation.md) | ||
| * [LLM use in evaluation](llm-use-in-evaluation.md) | ||
| * [Custom evaluation (custom metrics)](custom-evaluation.md) | ||
| * [Input](input.md) | ||
| * [Output](output.md) |
There was a problem hiding this comment.
These links don't work. If we want them to be relative, they should start with docs/. For example, docs/metrics.md. However, I think in order for the links to work on PYPI page https://pypi.org/project/graphrag-eval/ we should make them absolute, like so https://github.com/Ontotext-AD/graphrag-eval/blob/main/docs/metrics.md
There was a problem hiding this comment.
Also "Configuration" links to a non existing md file documentation.md, it must link to the configuration.md
There was a problem hiding this comment.
Fixed to relative links.
Relative links are shorter to read and edit, and work in an editor and a downloaded folder.
How useful is it for the documentation to be accessible from the PyPi page?
docs/installation.md
Outdated
| ``` | ||
|
|
||
| To evaluate `answer_relevance` and answer correctness metrics (`answer_recall`, | ||
| `answer_precision`, `answer_f1`; see section [Output keys](output.md)) or |
There was a problem hiding this comment.
| `answer_precision`, `answer_f1`; see section [Output keys](output.md)) or | |
| `answer_precision`, `answer_f1) see section [Output keys](output.md) or |
There was a problem hiding this comment.
This suggestion fixes the formatting, but I also suggest to link to the metrics file instead of the output and define how the metrics are calculated in the metrics file
There was a problem hiding this comment.
Why the suggestion? We have ( ... ; ... ).
docs/usage.md
Outdated
| ``` | ||
|
|
||
| `evaluation_results` is a list of statistics for each question, as in section | ||
| [Example wvaluation results](examples/output.md). The format is explained in section |
There was a problem hiding this comment.
| [Example wvaluation results](examples/output.md). The format is explained in section | |
| [Example evaluation results](examples/output.md). The format is explained in section |
There was a problem hiding this comment.
Fixed, thanks. I should re-read, run a spell checker and maybe a LLM.
docs/usage.md
Outdated
| [Example wvaluation results](examples/output.md). The format is explained in section | ||
| [Output Keys](output.md) | ||
|
|
||
| # Command Line Use |
There was a problem hiding this comment.
| # Command Line Use | |
| ## Command Line Use |
There was a problem hiding this comment.
Why? Should we define top-level heading "Usage" and level-2 headings "Use as a library" and "Command-line use"?
docs/usage.md
Outdated
| reference dataset and `actual_answer` in the target data to evaluate | ||
| 1. For custom metrics: | ||
| 1. Define the metrics in the [configuration file](configuration.md) | ||
| 1. Include refernce and target inputs used by the metrics |
There was a problem hiding this comment.
| 1. Include refernce and target inputs used by the metrics | |
| 1. Include reference and target inputs used by the metrics |
docs/usage.md
Outdated
| 1. For `answer_relevance`, include `actual_answer` in the reference | ||
| dataset | ||
| 1. For answer correctness metrics (section | ||
| [Output](output.md)), include `reference_answer` in the | ||
| reference dataset and `actual_answer` in the target data to evaluate | ||
| 1. For custom metrics: |
There was a problem hiding this comment.
This is inconsistent, for the answer relevance and for the custom metrics we don't link to the definition, but for answer correctness we do. Also, I would define the metrics (except for the custom metrics) in the metrics file, and link to it.
There was a problem hiding this comment.
Added links and made other changes to this section. Linked to metrics here.
As for changes to the contents of metrics.md in comments to that file.
Replaced the keyword by words without formatting, because a keyword suggests that the value is a key in the output.
|
@nelly-hateva please close threads. |
docs/. Benefits:Tests
README.mdanddocs/*TODO/questions: